MiniMax M3 428B · Speculative decoding
B200 FP4: M3 EAGLE vs Off speculative decoding
Speculative decoding comparison of M3 EAGLE versus Off on B200 FP4 (NVIDIA Blackwell) running MiniMax M3 428B. Throughput, cost, and interactivity differences across LLM workloads. Use the chart controls below to switch sequences and metrics — same interactions as the main inference chart.
MTP acceptance-rate implementations differ across inference engines. Points from different engines are not directly comparable on the same curve — throughput and cost at matched interactivity may reflect engine-level differences rather than pure speculative decoding gains. Interpret cross-engine comparisons with caution.
Throughput at 71 tok/s/user on MiniMax M3 428B (B200 FP4): M3 EAGLE hits 8,846 tok/s/chip, Off hits 8,368. Costs per million tokens land at $0.054 and $0.057 respectively. M3 EAGLE delivers 6% more tok/s/chip on the same chip, which makes it 5% cheaper per token. Speculative decoding trades extra compute on draft tokens for fewer decoding steps — the payoff depends on sequence length and batch size.
Around the middle of the 22–220 tok/s/user interactivity band, at 121 tok/s/user on MiniMax M3 428B (B200 FP4): M3 EAGLE runs 5,814 tok/s/chip at $0.083 per million tokens, Off runs 4,517 at $0.11. M3 EAGLE delivers 29% more tok/s/chip on the same chip, which makes it 22% cheaper per token. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less.
At 171 tok/s/user on MiniMax M3 428B (B200 FP4), M3 EAGLE delivers 5,229 tok/s/chip at $0.092 per million tokens; Off delivers 1,138 tok/s/chip at $0.42. M3 EAGLE delivers 4.6× the tok/s/chip on the same chip, which makes it 78% cheaper per token. Speculative decoding accepts draft tokens to reduce per-token latency — gains vary by workload and prompt distribution. (Numbers reflect this URL's pinned 8k/1k · fp4 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | M3 EAGLE8,846Off8,368 | M3 EAGLE5,814Off4,517 | M3 EAGLE5,229Off1,138 |
| Cost ($/M tok) | M3 EAGLE$0.054Off$0.057 | M3 EAGLE$0.083Off$0.11 | M3 EAGLE$0.092Off$0.42 |
| tok/s/MW | M3 EAGLE5,172,906Off4,893,496 | M3 EAGLE3,399,910Off2,641,510 | M3 EAGLE3,057,957Off665,389 |
| Concurrency | M3 EAGLE~33Off~232 | M3 EAGLE~20Off~64 | M3 EAGLE~70Off~3 |
